Beyond monogenetic rare variants: tackling the low rate of genetic diagnoses in predominantly antibody deficiency

Abstract

Predominantly antibody deficiency (PAD) is the most prevalent form of primary immunodeficiency, and is characterized by broad clinical, immunological and genetic heterogeneity. Utilizing the current gold standard of whole exome sequencing for diagnosis, pathogenic gene variants are only identified in less than 20% of patients. While elucidation of the causal genes underlying PAD has provided many insights into the cellular and molecular mechanisms underpinning disease pathogenesis, many other genes may remain as yet undefined to enable definitive diagnosis, prognostic monitoring and targeted therapy of patients. Considering that many patients display a relatively late onset of disease presentation in their 2^nd or 3^rd decade of life, it is questionable whether a single genetic lesion underlies disease in all patients. Potentially, combined effects of other gene variants and/or non-genetic factors, including specific infections can drive disease presentation. In this review, we define (1) the clinical and immunological variability of PAD, (2) consider how genetic defects identified in PAD have given insight into B-cell immunobiology, (3) address recent technological advances in genomics and the challenges associated with identifying causal variants, and (4) discuss how functional validation of variants of unknown significance could potentially be translated into increased diagnostic rates, improved prognostic monitoring and personalized medicine for PAD patients. A multidisciplinary approach will be the key to curtailing the early mortality and high morbidity rates in this immune disorder.

Introduction

Primary immunodeficiencies (PIDs) are rare inherited disorders that impair the human immune system. Also referred to as inborn errors of immunity, these disorders manifest as increased susceptibility to infections, autoimmunity, gastrointestinal disease, allergy, and/or malignancy. Long considered to be rare (1/10,000 live births),^1–3 it has emerged that disease incidence may be higher (1/1000 to 1/5000) than previously thought.⁴ This increase likely reflects improved detection as a result of improved awareness and mechanistic understanding of these disorders with >400 causative gene defects now identified.⁴ The rapid increase in disease identification over the past decade is driven by advances in genomic sequencing, functional studies and improved definition of clinical phenotypes.^4,5

The diversity of PID is captured by nine categories that are defined based on the clinical, genetic and immunological characteristics of disease.^4,5 Of these nine categories, predominantly antibody deficiency (PAD) has the highest disease prevalence (1/25,000).^6–8 Despite advances in genomics, the diagnostic rate for PAD remains low (<20% of cases), especially in nonconsanguineous populations.^6,9,10 Prior to genomics approaches, inherited mutations in a restricted number of genes were specifically associated with a PAD phenotype (Fig. 1; Table 1).^9,11 Increased genomic analysis of PAD patients has identified variants in genes associated with other PID phenotypes (Fig. 1; Table 1), thereby highlighting the complex genetics underpinning this disease group.^9,10 These developments change the view on the patient group currently defined as PAD in that it might harbor multiple forms of PID with the common feature of antibody deficiency. Proper stratification of PAD patients could aid in curtailing the early mortality and high morbidity rates in this population.

Fig. 1.

Timeline detailing the single gene variants identified in PAD patients. For reference, key events advancing the field of genomics have also been added. ^$Identified by pulse-field electrophoresis, ^*date of publication, ^{^}AR and AD mutations identified in the same gene underlying different clinical phenotypes, ^#AR and AD mutations identified underlying similar clinical phenotypes, ^&date genetic mutations identified in PAD, identified previously in other clinical phenotypes (See Table 1)

Table 1.

Monogenic lesions identified in PAD patients

Decade	X-linked		Autosomal recessive		Autosomal dominant			References
	Genes	n	Genes	n	Genes	n	%
1980s	–	0	IGKC	1	–	0	0	272
1990s	BTK, CD40LG	2	IGHM, IGLL1, CD79A, BLNK, IGHC deletions	5	–	0	0	20–23,273
2000s	–	0	AICDAa, UNG, TNFRSF13Ba, CD19, CD79B, MSH6, TNFRSF13C, CD40, ICOS	9	AICDAa, TNFRSF13Ba	2	18	24,35,81,95,105,109,110,111,113,118,274
2010s	ATP6AP1, SH3KBP1	2	CD81, CD20, PIK3R1b, CD21, PIK3CDb, MOGS, INO80, RAC2, TCF3a, SLC39A7, ARHGEF1, TRNT1, LRBA, CD27, PRKCD, IL21R, IL21, ICOSL, TNFSF13	19	IKZF1, CARD11, TCF3a, PIK3CDb, TNFSF12, NFKB2, PIK3R1b, NFKB1, IRF2BP2, SEC61A1, PTEN, TOP2B, CTLA4, PLCG2	14	40	17,19,25–28,33,55,75,83,88,91–94,96,97,108,112,123,124,127,129,137,140,145,154,156,157,217,239,241–243,275–281

^aAR and AD mutations identified underlying similar clinical phenotypes

^bAR and AD mutations identified in the same gene underlying different clinical phenotypes

Clinical heterogeneity in patients with predominantly antibody deficiency

PAD patients typically suffer from recurrent sinopulmonary infections, which may be severe and persistent, with pronounced antibody deficiency and impaired vaccination responses underpinned by defects in B-cell development, maturation and/or function (Fig. 2). Treatment is centered around the humoral immune defect and includes lifelong immunoglobulin replacement therapy (IgRT) and prophylactic antibiotics to reduce infection.^7,8

Fig. 2.

Schematic representation of the impact of gene variants on B cell development in bone marrow and peripheral lymphoid organs. The checkpoint processes that a B cell is required to pass are indicated by letters A to K. The boxes below indicate where genetic variants impair one of these processes causing a block in B-cell development, maturation or function. When inherited in an autosomal dominant fashion, the gene variants colored in red cause defects in a process distinct from those abrogated when inherited in an autosomal recessive manner in blue. BCR B cell receptor, TF transcription factors, TI T cell-independent, TD T cell-dependent, Ig CSR immunoglobulin class switch recombination

Agammaglobulinemia is the archetypal PAD,⁸ characterized by complete or near absence of B cells (B cells <1%) and absence of all major serum immunoglobulin isotypes (IgG, IgM, IgA).⁵ Up to 85% of agammaglobulinemia patients are males with X-linked (XL) mutations in Bruton’s tyrosine kinase (BTK) gene.^12–14 BTK has a critical role in precursor B-cell development in bone marrow.¹⁵ The remaining 15% of patients have autosomal recessive (AR) or -dominant (AD) agammaglobulinemia as a result of mutations in genes responsible for the formation of the pre-B cell receptor complex or transcription factors pivotal for early B-cell development.^16–36

Other forms of PAD are less well defined mechanistically, and subclassification of these is based on clinical parameters. Combined variable immunodeficiency (CVID),^6,11 is the most clearly defined form on the basis of recurrent sinopulmonary infections, poor vaccination responses and reduced serum IgG coincident with reduced IgA and/or IgM levels.^11,37–40 Diagnostic criteria for CVID are continuously evolving to represent the diversity of clinical and immunological phenotypes.^41–43 Patients with recurrent infections, impaired vaccination responses and reduced IgG in the context of normal IgA and IgM are defined as (unspecified) hypogammaglobulinemia.³⁹ Individuals with reduced or absent IgA but normal levels of other isotypes and specific antibodies are defined as selective IgA deficiency, a condition that is frequently asymptomatic.^5,44 Furthermore, there are patients with an IgG subclass deficiency (IgSCD) in the context of normal total IgG, IgA and IgM and patients with impaired polysaccharide antibody responses in the context of normal serum immunoglobulins (specific antibody deficiency; SpAD).^45–47 The clinical severity of these different syndromes, for the most part, is dependent on the extent of abnormal serological findings.^5,39

IgRT and antibiotics are the mainstay of therapy for PAD, and in the vast majority of patients, these successfully reduce the infectious burden.^48,49 However, 50–68% of patients develop non-infectious complications (NICs) that include autoimmune disease, gastrointestinal disease and malignancy.^7,48,50–52 These NICs are not ameliorated by IgRT or antibiotics and thus are the predominant cause of morbidity and early mortality in PAD patients.^48,51–53 Importantly, NICs appear to be driven by immune dysregulation beyond B-cell and antibody defects. In fact, studies from multiple cohorts around the world have identified key differences in B-, T- and innate cell numbers and maturation stages associated with NICs in PAD.^{45,48,50,54–70} Furthermore, evidence is emerging that CVID-associated NICs are driven at least in part by IFN-γ.^65,71,72 Therefore, affected patients may benefit from therapeutic strategies that specifically target the particular immune cells or molecules that drive the disease. The complexity of antibody deficiency syndromes in combination with the high diversity of NICs highlights the requirement for more precise diagnosis, better prognostic markers and targeted therapeutics to more effectively manage disease.

Genetic defects underlying impaired B-cell development and antibody responses in PAD

In those PAD cases where causative gene defects have been identified (Fig. 1), detailed analysis of B-cell development and function has uncovered the critical role of these genes in B-cell immunobiology (Fig. 2). Progenitor B-cell development and mature B-cell responses take place sequentially in the bone marrow and secondary lymphoid tissues. Mature B-cell responses culminate in the generation of high-affinity memory B cells and antibody-producing plasma cells specific to invading pathogens.⁷³ Critical processes and developmental checkpoints in progenitor B cells in bone marrow have been shown to be affected by inherited mutations. Of these, germline biallelic loss-of-function (LOF) mutations in TOP2B, the gene encoding DNA topoisomerase 2-beta, impair common lymphoid progenitor progression to the pro-B-cell stage (Fig. 2).^16,26 This defect occurs prior to initiation of V(D)J recombination, a shared process in thymocytes and progenitor B cells, which, when impaired, leads to severe combined immunodeficiency. V(D)J recombination of the IGH locus in pre-BI cells is critical for the formation of the pre-B-cell receptor (BCR), which upon expression signals cell proliferation, survival and developmental progression into the pre-BII cell stage.⁷⁴ AD LOF mutations in the gene encoding the transcription factor Ikaros (IKZF1), as well as AD and biallelic AR mutations in E47 (TCF3), result in substantial decreases in precursor B cells due to the pivotal role these transcription factors play in B-cell development (Fig. 2).^17–19,75 Furthermore, genetic defects in pre-BCR component-encoding genes (IGHM, IGLL1, CD79A and CD79B),^{20–22,24,35} those coding for downstream signaling components (BTK, BLNK, PIK3CD, PIK3R1, RAC2) (Fig. 3),^{12,14,23,27,28,30–34,36,76,77} SLC39A7 encoding zinc transporter ZIP7,²⁵ and enzyme TRNT1, which is responsible for synthesizing CCA sequences at the 3’ end of tRNA,^78,79 impair developmental progression (Fig. 2), resulting in absent or markedly reduced B-cell numbers and agammaglobulinemia.

Fig. 3.

Schematic of proteins encoded by causal genes identified in PAD (genes associated with the PAD phenotype, red; and a phenotype other than PAD, blue) and their role in signaling  of B-cell surface receptors, including BCR. Only the most important interacting proteins and pathways are illustrated

Following initial proliferation, pre-BII cells return to a resting phase, during which Ig light chain gene rearrangements are induced to generate the complete BCR. Once complete, these cells develop into immature B cells that undergo selection, and only those with a BCR displaying low affinity for self-antigen will egress to peripheral lymphoid organs as transitional B cells. To date, no genetic defects have been shown to impair central tolerance at the immature B-cell stage other than those genes affecting BCR signaling that already affect pre-BII cells (Fig. 2). However, activation-induced cytidine deaminase (AID), which impacts later-stage antigen-dependent B-cell receptor maturation, has been implicated to adversely impact central tolerance mechanisms.⁸⁰ Transitional B cells require two survival signals, one through tonic signaling from the BCR and the other from engagement of soluble B-cell activating factor (BAFF) with its cell surface-expressed BAFF receptor (BAFF-R). Naive mature B-cell survival is impaired in patients with biallelic LOF mutations in TNFRSF13C, encoding BAFF-R,^9,81,82 and in patients with AD LOF mutations in TNF-related weak inducer of apoptosis (TWEAK) encoding TWEAK (Fig. 2). TNFRSF13C LOF mutations result in complete absence of BAFF-R expression on the B-cell surface,^81,82 whereas TWEAK mutations result in soluble mutant TWEAK protein binding to BAFF in serum (Fig. 3). As a result, BAFF binding to BAFF-R is reduced, thereby abrogating BAFF-mediated activation of the noncanonical NFκB (nuclear factor kappa-light-chain-enhancer of activated B cells) pathway, which is required for B-cell survival, proliferation and T-cell independent (TI) isotype switching.⁸³

Each naive mature Bcell expresses a BCR with a unique specificity and the potential to recognize a foreign antigen. The strength of the BCR-antigen interaction and the engagement of co-receptors dictates the level of B-cell activation. Upon engagement of the BCR by its cognate antigen, the BCR translocates to lipid rafts, which put the BCR in close contact with a co-receptor complex consisting of CD19, CD21, CD81, and CD225.⁸⁴ This CD19-complex functions to reduce the antigen-dependent signaling threshold through the intracellular domain of CD19 that engages phosphatidylinositol-3-kinase (PI3K) and amplifies NFκB signaling (Fig. 3).^85,86 The CD19 complex can also be engaged through the complement pathway because CD21 is a receptor for complement fragment C3d. C3d-loaded antigens can be bound simultaneously to BCR and CD21 (Fig. 3), thereby mediating co-localization of the BCR and CD19-complex. Biallelic LOF mutations in CD19, CD21, and CD81 disrupt the function of the co-receptor complex (Fig. 2). Mutations in CD19 and CD81 result in the absence of expression of CD19 and intracellular signaling and subsequently a severe reduction in serum IgG, with normal/low IgA or IgM,^87–90 whereas CD21 mutations result in a milder phenotype as these only impair complement-dependent interactions.^90,91 XL mutations in the accessory protein Ac45 encoded by ATP6AP1, part of the vacuolar H⁺-ATPase complex, which is responsible for acidification of cellular compartments, including lysosomes, have been shown to impact B-cell activation (Fig. 2). However, the precise mechanism has yet to be fully elucidated.⁹² Finally, biallelic LOF mutations in the gene encoding SH3 domain-containing kinase-binding protein 1 (SH3KBP1), which associates with BLNK (B-cell linker protein) upon BCR ligation (Fig. 3), result in intrinsic defects in BCR-mediated activation (Fig. 2). As this protein operates downstream of active rather than tonic signaling, which occurs at the pre-BCR stage,⁹³ maturation rather than development of B cells is impacted in SH3KBP1-deficient individuals, resulting in normal numbers of all major B-cell subpopulations.⁹⁴ Together, mutations in the aforementioned genes result in decreased BCR-mediated signaling and reduced serum IgG, IgM and/or IgA.^{87–91,94,95}

The activation signal provided by antigen-BCR interactions and amplified by the CD19 complex is still insufficient for B-cell expansion and differentiation. This requires an additional signal, which is typically provided by CD4⁺ T cells but can be generated through other means, such as signaling through pattern-recognition receptors including Toll-like receptors (TLRs) or via extensive BCR cross-linking. T-cell independent (TI) B-cell responses are impaired in patients with biallelic LOF mutations in MS4A1 encoding CD20⁹⁶ or with AD gain-of-function (GOF) mutations in the gene encoding caspase recruitment domain-containing protein 11 (CARD11) (Fig. 2).⁹⁷ Normally, CD20 forms a homo-oligomer, crosslinked by TI antigens which regulate transmembrane calcium transport (Fig. 3).^98,99 Therefore, CD20 deficiency results in impaired responses to TI antigens and reduced SHM in IgG heavy chains.⁹⁶ In contrast, CARD11 is an important adapter protein downstream of protein kinase C that forms a complex with BCL-10 and MALT1 to activate the canonical NFκB pathway (Fig. 3). GOF mutations in CARD11 result in increased complex formation and hence hyperactivation of the canonical NFκB pathway in B cells. The perturbed NFκB response results in reduced responses to TI antigens, deficiencies in class switched memory B cells, reduced Ig levels in serum and poor plasma cell generation in vitro.⁹⁷

CD4⁺ T-cell co-stimulation is provided in an antigen-specific manner when B cells present cognate antigen in the context of MHC class II molecules. This occurs following initial activation of CD4⁺ T cells by recognition of cognate antigen presented by dendritic cells (DCs). Activated CD4⁺ T cells (namely, T follicular helper cells; TFH) upregulate CD154 (also known as CD40L) and the inducible costimulator (ICOS) that bind their ligands (CD40 and ICOSL, respectively) on B cells. These receptors provide important signals for B-cell activation and BCR maturation. T-cell-dependent (TD) responses induce germinal centers (GCs) consisting of highly proliferative B cells and follicular DCs (FDCs). FDCs are of stromal origin and as such express molecules crucial for B-cell survival and proliferation. In particular, FDCs present C3d-loaded antigen to B cells engaging BCR and CD21 on their cell surface. This selectively provides survival signals to high-affinity B cells. Together, the signals provided by FDC and activated CD4⁺ T cells are crucial for B-cell affinity maturation by inducing mutations in the variable regions of Ig heavy and light chain genes of proliferating B cells.¹⁰⁰ These somatic hypermutations (SHM) are induced by AID, a direct target of CD40 signaling. In addition, AID induces lesions in the Ig switch regions that are located upstream of all Ig constant gene regions, and these mutations mediate Ig class switch recombination (CSR). As a result, the IgM isotype of the BCR can be replaced by IgG, IgA, or IgE. Ultimately, activated B cells are selected for high-affinity antigen binding and differentiate into memory B cells and Ig-producing plasma cells.^73,101

LOF mutations in 4 genes have been shown to cause PAD due to impaired T-cell help to B cells. XL LOF mutations in CD40LG (encoding CD154)¹⁰² and biallelic LOF mutations in ICOS, ICOSL, and CD40 impair T-cell-dependent B-cell activation (Fig. 2).^102–108 Because CD40–CD154 interactions directly induce AID expression, patients with mutations in these genes have an Ig CSR defect and were previously defined as having hyper-IgM syndrome.^102,103 CD40-CD154 and ICOS-ICOSL interactions are important for T-cell function beyond B-cell help. Disruptions of these pathways result in both cellular and humoral immune defects, and patients with affected variants are therefore classified as having a combined immunodeficiency.⁴

Ig CSR deficiencies result from null mutations in AICDA encoding AID, which abolish CSR and SHM.¹⁰⁹ In addition, CSR deficiencies can be caused by mutations in UNG¹¹⁰ and MSH6,¹¹¹ which both function in the repair of double-stranded DNA breaks, as well as mutations in the chromatin remodeler INO80, which causes impaired INO80-complex function (Fig. 2).¹¹² AR or AD mutations in TNFRSF13B encoding TACI (transmembrane activator and calcium-modulator and cyclophilin ligand interactor) result in reduced binding of TACI to its ligands, BAFF and APRIL (a proliferation-inducing ligand).^113–119 Ligation of TACI has been suggested to support both TD and TI responses, and loss of TACI function impairs the B-cell response.¹²⁰

Downstream of the BCR and costimulation molecules, the intracellular modulation of B-cell signaling (Fig. 3) plays a pivotal role in cell maturation and function. In particular, examination of cells from patients with PID has uncovered that tight regulation of both the PI3K and NFκB pathways is required for B- and T-cell homeostasis, with dysregulation shown to profoundly impact immune cell function, causing antibody deficiency and immune dysregulation.^121–126 AD LOF mutations in ARHGEF1 encoding Rho guanine nucleotide exchange factor 1,^127,128 PIK3R1, encoding p85α (the regulatory subunit of PI3K),^123,124 PTEN, encoding phosphatase and tensin homolog,^129,130 or GOF mutations in PIK3CD, encoding p110δ (the catalytic subunit of PI3K),^125,126 result in unstrained PI3K/AKT signaling. LOF mutations in PIK3R1 and GOF mutations in PIK3CD result in conformational changes in either p85α or 110δ collectively, resulting in disrupted binding and thus regulation of p110δ kinase activity by p85α.^{123–126,131} In contrast, LOF mutations in PTEN impair PTEN inhibition of PI3K activity.¹²⁹ Alternatively, ARHGEF1 mutations cause disturbed regulation of downstream targets of ARHGEF1, including RhoA, resulting in impaired restraint of AKT activation.¹²⁷ These defects result in reduced naive and/or increased transitional B cells, with reduced marginal zone and memory B cells.^123–129

Crosstalk between the canonical and noncanonical NFκB pathways modulates immune cell activity.¹²¹ Haploinsufficiency of NFKB1, which encodes p50, impairs signaling through the canonical NFκB pathway.^{121,132–134} In contrast, AD LOF mutations in NFKB2 affect both the canonical and noncanonical NFκB pathways.^9,121,135 In both genetic disorders, B-cell maturation is disturbed (Fig. 2) with severely reduced serum IgG and IgA, Ig class-switched memory B-cell numbers and antigen-specific antibody responses.^{9,132,134–137} Serum IgM is differentially affected with normal or high levels in AD GOF PIK3CD and AD LOF PIK3R1 and normal or low levels in AD LOF PTEN, ARHGEF1, NFKB1, and NFKB2 defects.^{4,9,29,87,102–113,115–119,121–129,132,134–149} In addition to B-cell intrinsic defects in patients with dysregulated PI3K or NFκB pathways, defects in other immune compartments, including innate cells and T cells, may influence B-cell homeostasis.^121,122 This includes alterations in the number and function of Tfh cells, which could directly impact the T-cell help required for memory B-cell and plasma cell generation.^{121,122,149,150} Furthermore, altered innate cell function^121,151,152 may directly affect B-cell functionality, although the exact nature of this will need to be further elucidated.

Terminally differentiated B cells or plasma cells (Fig. 2) produce high quantities of antigen-specific antibodies for the clearance of infectious agents.¹⁵³ A range of gene variants have been identified to underlie for plasma cell deficiency and reduced antibody production, including: AD mutations in SEC61A1 encoding the Sec61 complex crucial for transporting proteins from the cytosol to the endoplasmic reticulum;^154,155 IRF2BP2 encoding interferon factor 2-binding protein-2, which is important in regulating type I interferon-dependent transcription;¹⁵⁶ biallelic LOF mutations in TNFSF13 encoding APRIL that is known to bind TACI and B-cell maturation antigen (BCMA), which is required for long-term B-cell memory maintenance (Fig. 3);¹⁵⁷ as well as AR mutations in MOGS encoding mannosyl-oligosaccharide glucosidase, an enzyme required for N-glycan trimming.¹⁵⁸ These defects do not affect total B-cell numbers (>1%)⁵ but result in plasma cell deficiency and reduced antibody production (Fig. 2).^154–157

The identification of causal gene mutations in PAD has enabled the elucidation of their key roles in pre-BCR formation, B-cell survival and activation, TD and TI B-cell responses, Ig CSR and plasma cell formation (Fig. 2). Regardless of whether these defects occur early or late in the development process, they all culminate in the absence of functional antibody generation, thereby rendering patients susceptible to recurrent infections. It is also important to note that some of these genetic defects may also underlie coexistent NIC. Further analysis of their role in driving disease in this context is required.

Successes and new challenges for genomic approaches in diagnosis of PAD

Overview of approaches for genetic analysis

Identification of a causative variant is regarded as a definitive diagnosis of PID. DNA sequencing developed in the 1970s (Fig. 1), especially the Sanger method, has enabled targeted sequencing of the candidate gene region.¹⁰ However, it was only after 1993, when the human genome sequence became available, that the first PAD-causing gene was identified. Forty years after the initial clinical description of XLA,¹⁵⁹ it was shown that genetic defects in BTK were the culprit (Fig. 1).¹⁴

Clear phenotype-genotype associations are rare in PID due to the heterogeneous and complicated clinical, immunological, and genetic nature of disease. For this reason, the restricted number of candidate genes that can be analyzed simultaneously or consecutively with Sanger sequencing makes it expensive and laborious for variant identification.¹⁶⁰ However, despite this drawback, due to its high sensitivity and accuracy, the Sanger method remains the gold standard for validation of variants identified by next-generation sequencing (NGS) methodologies.^{2,10,161–163}

The completion of the Human Genome Project in the early 2000s,^164–167 and subsequent advancements in NGS technologies, including targeted sequencing panels, whole exome (WES) and whole genome (WGS) sequencing, have provided: 1. cost-effective and efficient high-throughput sequencing technologies and analytical tools, 2. unbiased approaches for interrogation of the human genome superseding Sanger sequencing, and 3. open access data sharing and software tools. All of these technologies have significantly contributed to the increased discovery of novel causal genes in PID (Table 1; Fig. 1).^10,160,168 Targeted sequencing panels are rationally designed on the basis of genes implicated in PID or with a known function in the immune system.^10,169–172 However, by definition, such panels have a limited capacity to drive discovery of new causal gene variants.¹⁶⁰ In contrast, WES targets all protein-encoding regions, encompassing ~1.5% of the total human genome sequence, and has a higher yield in detecting causal mutations compared to Sanger sequencing and targeted sequencing panels.^10,160,173 It has been reported that up to 85% of disease-causing variants occur within coding regions.^10,173,174 Nevertheless, this suggests that more attention should be focused on the identification of variants in non-coding regions, which may still play a role in the regulation of gene expression and function.^175,176 However, a caveat of WES is the inability to detect mutations in non-coding regions, guanosine-cytosine-rich regions and sequences shared by pseudogenes, as well as the limited capacity to detect structural variants including alterations in copy number, all of which are readily identifiable by using WGS.^10,160,177 On the other hand, WGS generates ~20× larger quantities of data than WES, which requires a higher computing capacity and additional data management considerations, including storage, safeguarding of data and, importantly, higher rates of as yet unclassifiable/unknown variants. Despite this, with ongoing technical development, it will only be a matter of time before WGS will replace WES as the gold standard for novel variant identification by genomics.^4,10,160

NGS by either WES or WGS has revolutionized genomics, resulting in increased gene discovery. Prior to the year 2000, only eight PAD causal genes had been identified, with the majority inherited in an AR fashion (n = 6) (Fig. 1; Table 1). In the period post-2000, the first AD lesions were identified with AR lesions predominating over AD and XL variants. In contrast, post-2010, a dramatic increase in the identification of both AR and AD variants occurred with almost equal distribution of variants between AR and AD. Over the entire 20 years preceding completion of the Human Genome Project, only two XL disorders were identified. Overall, a dramatic increase (85.1%) in the monogenic lesions underlying PAD has been facilitated by the Human Genome Project and NGS (Fig. 1; Table 1).

Pipeline for identification of causal variants

The diagnostic challenge in WES and WGS is the selection of a causative variant from thousands of NGS reads (up to 50,000 per exome for WES) and hundreds of variants within the standard human genome sequence.^161,177,178 The large number of gene variants identified need to be filtered (Fig. 4), a process typically done on the basis of available information, as discussed below.¹⁷⁹

Fig. 4.

Schematic representation of the pipeline for the identification and categorization of rare variants from NGS. Based on refs. ^{160,161,179,190,271}

Variants are assessed on the basis of their allele frequency in the general population using population databases including 1000 Genomes and gnomAD,^180,181 with an allele frequency of <0.01 defined as rare.^182,183 Variants above this threshold are excluded because they are unlikely to represent a molecular cause of a rare disease such as PAD.^{160,161,182,183} After filtering, the remaining variants are classified according to population data, allelic distribution data, variant-based computational data, and functional and biological data based on the American College of Medical Genetics and Genomics (ACMG) guidelines for determining variant pathogenicity (Fig. 4).¹⁷⁹

Subsequently, the functional impact of a variant is predicted using publicly available algorithms to predict pathogenicity in silico including sorting intolerant from tolerant (SIFT), polymorphism phenotyping v2 (PolyPhen2), and combined annotation-dependent depletion (CADD) (Fig. 3).^181,184,185 Variants resulting in a premature stop codon or deletion, for example, strongly predict pathogenicity. Missense mutations may be pathogenic but require further evidence to confirm pathogenicity (Table 2). Conversely, the absence of a “predicted” effect may indicate that the variant is benign. However, the current in silico tools and available data can yield conflicting information for some variants, such that the variant is predicted as benign by one tool but as likely pathogenic by another. These variants are termed variants of unknown significance (VUS) (Table 2). Consideration of the known information about the gene product and the predicted impact on protein function can indicate whether a VUS may be pathogenic.¹⁷⁹ In gene discovery, inheritance patterns can indicate the type of defects that can be expected, i.e., AR, XL, or AD. The mechanisms underlying AD diseases can be haploinsufficiency, GOF or dominant negative activity of the mutant allele,^186,187 thereby interfering with the function of the wild-type allele.^75,188 In contrast, AR inheritance is the result of biallelic variants, either homozygous or compound heterozygous, which typically result in LOF of the encoded protein.^186,187 These concepts should be placed in the context of known information regarding the mutated gene. For example, if biallelic variants in a known causal gene underlie disease, identification of a heterozygous variant in the same gene should be functionally validated to prove or disprove pathogenicity. Both monoallelic and biallelic variants have been identified in numerous genes underlying PAD, including AICDA, TCF3, PIK3CD, and PIK3R1 (Fig. 1; Table 1),¹⁸⁹ therefore, functional analysis is required to prove pathogenicity versus carrier status. In addition, databases of pathogenic mutations including the Human Gene Database, can be utilized to establish whether a variant has been previously identified in the gene of interest, strengthening the evidence for variant pathogenicity (Fig. 4).^180,181

Table 2.

Categorization of rare gene variants identified using NGS

Category	Predicted variant pathogenicity	Examples of variants and their effects
1	Benign	– Minor allele frequency too high for disorder
		– Does not segregate with disease
		– Synonymous mutation, i.e. no change in amino acid sequence
		– No deleterious effect demonstrated in functional studies
2	Likely benign	– Observed in cis with pathogenic variant in the same gene, or in trans with AD pathogenic variant
		– Computational evidence argues against deleterious effect
3	Uncertain significance	– Evidence supporting benign or pathogenic nature of variant is contradictory, e.g. novel nucleotide variant results in an amino acid change in an unknown gene
4	Likely pathogenic	– Novel missense variant at same amino acid residue as previously identified pathogenic missense variant or variant located at a mutational hot spot and/or critical functional domain where no benign variants have been identified
		– De novo mutation (i.e. confirmed parents both do not carry variant)
5	Pathogenic	– Predicted null variant (e.g. nonsense, frameshift, stop codon, splice-site, single or multicodon deletion) in gene with proven deleterious effect
		– Variant has been previously reported to be deleterious in this disease

Based on refs. ^160,179

The information from the aforementioned processes enables stratification of variants based upon predicted pathogenicity: benign or likely benign (categories 1 and 2), VUS (category 3), likely pathogenic (category 4) and pathogenic (category 5) (Fig. 4; Table 2).^179,190 Benign and likely benign (categories 1 and 2) variants are disregarded, whereas pathogenic (category 5) variants are given a confirmatory diagnosis (Fig. 4). The major challenge is for VUS (category 3) or likely pathogenic variants (category 4), which are reported in the majority of patients. Recently, in patients with PAD, variants have been identified in genes not previously associated with these disorders, including PRKDC, MAPK8, PRRC2A, TNIP1, and BCL2L1. The majority of these genes have been shown to impact B-cell function.^191–196 Therefore, functional validation will be required to prove causality (Fig. 4). Furthermore, variants in genes such as CD27, XIAP, LRBA, and CTLA4^{4,5,10,196,197} have been identified in patients previously diagnosed with PAD but have since been reclassified into separate PID categories. Thus, due to the evolving patient genotype-phenotype association, functional validation may facilitate deeper understanding of the implicated gene and potentially illuminate immunological concepts that are yet to develop, which informs patient reclassification and solidifies the diagnosis. Despite the enormous advancements in NGS applications in the past decade,^17–19,26 over 80% of PAD patients remain genetically undiagnosed.^6,9,193,196 This may suggest that there are many more genes to be identified that, when mutated, will impair antibody responses. Alternatively, the heterogeneous presentations of several monogenic conditions, including CTLA-4 and NFκB1 haploinsufficiency in terms of age of onset, severity of disease and penetrance,^133,198 suggest that other mechanisms including epigenetics, may play a crucial role in driving disease.

Alternative genetic variants underlying PAD

Although genomics has focused on the identification of small lesions, including single nucleotide variants and indels (insertions/deletions), it is becoming more evident that gross defects as well as disease-modifier genes may contribute to the PAD phenotype.

Structural variants

Traditionally, genomics has focused on the identification of small lesions. By definition, structural variants are large genomic alterations (>50 bp) including deletions or duplications, inversions and translocations, which have the capacity to disrupt gene function and regulation and are termed copy number variants (CNVs) and balanced rearrangements, respectively.¹⁹⁹ Multiple genes with a critical role in the immune system are susceptible to large genetic alterations.^36,175

To date, few studies have identified structural variants in PID patients using WES due to the limited capacity of this method to detect these variants.¹⁶⁰ However, CNVs have been identified in PID patients using Sanger sequencing,³⁶ microarray analysis^{175,193,200,201} and WGS.¹⁷⁶ One study showed that CNV burden was higher in CVID patients than in healthy individuals, although the precise role of this in disease etiology has yet to be elucidated.²⁰⁰

Rare CNVs have been identified in PID genes including RAG1, WAS, NFKB1, XIAP, and LRBA, with pathogenicity proven by functional studies.^175,176,193 Although most CNVs were identified in coding regions, using WGS also identified CNVs in non-coding regions.¹⁷⁶ Structural variant calling of pathogenic CNVs may be important in certain PIDs. For example, causal variants in WAS are identified by traditional sequencing methods in 95% of Wiskott–Aldrich syndrome patients, with the remaining 5% representing structural variants that can only be detected with high reliability via WGS and comparative genomic hybridization arrays.¹⁷⁵

At present, accurate detection of structural variants is curtailed by the short-read lengths generated for WES and WGS. However, emerging long-read sequencing technologies show promise in improving detection accuracy. However, these methodologies are still under development and require further validation. Additionally, the decrease in cost and improvement in data handling and management will likely facilitate WGS overtaking WES in the genomic diagnosis of inherited diseases, including PID.^160,177

Taken together, wider adoption of WGS and integrative analytical pipelines for the identification of small (single nucleotide variants and indels) and gross (structural variants) lesions should be established. This will determine the contribution of structural variants to the genetic diversity of PAD and facilitate an increase in the identification of rare causal variants underlying disease, thereby increasing the rate of confirmatory diagnosis.

Common variants as disease modifiers in patients with PAD

In the early 2000s, genome-wide association studies identified short nucleotide polymorphisms (SNPs) associated with an increased risk of chronic immune-mediated diseases including autoimmunity, autoinflammation and inflammatory bowel disease.^202–204 Many of these SNPs are common (minor allele frequency; MAF > 1%)^182,183 and are typically regarded as a marker for an allele, rather than the genetic lesion causing the risk.²⁰⁵ However, some SNPs have been shown to alter gene function and could direct drug repurposing to alleviate disease.²⁰⁶

Common variants in MHC, ADAM, CD21, ICOS, and TNFRSF13B have been identified as susceptibility loci for CVID.^202–204 In particular, HLA-DQB1 confers genetic susceptibility to both IgA deficiency and CVID.²⁰⁷ In addition, heterozygous and homozygous variants in TNFRSF13B encoding TACI are more frequent in PAD patients (10%) than unaffected controls.^{113,115–119} However, heterozygous TNFRSF13B variants have been identified in asymptomatic relatives and healthy individuals, suggesting that this is a disease-modifying rather than a disease-causing gene for antibody deficiency.^{9,113,115–119}

The common variant c.1858C>T (p.R620W) in PTPN22 has the strongest association with autoimmunity outside of the MHC locus.^204,208 This mutation alters tyrosine phosphatase activity²⁰⁹ and is strongly associated with clinical autoimmunity in PAD.²¹⁰ It is plausible that common variants represent disease-modifying genes that contribute to heterogeneity of disease onset and clinical and immunological presentation in individuals with the same genotype. However, the prevalence and contribution of common variants to disease pathogenesis requires examination to determine whether they explain the conundrum of coexistent antibody deficiency and NIC. If sufficient evidence is uncovered, this should facilitate clinical reporting on the premise that these variants may represent biomarkers for prognostication and improved therapeutic targeting of PAD patients with NICs.

Here, we propose a polygenic model whereby the spectrum of immunological diseases is explained by varying contributions of genetic factors, including rare pathogenic variants and common variants, and non-genetic factors, including specific infections, the microbiome and diet. These factors have the capacity to impact epigenetic and transcriptional programs of immune cells. In this model, early-onset PIDs and chronic immune-mediated disease including autoimmunity, represent the two extremes of the disease spectrum. Early-onset PIDs are mostly driven by rare pathogenic variants, and chronic immune diseases are predominantly driven by common variants and non-genetic factors. In this context, we regard PAD disorders as an intermediate in this spectrum of diseases, whereby a rare pathogenic variant triggers or underlies the immune defects predisposing to disease susceptibility, with common variants and non-genetic factors acting as modifiers of the disease phenotype. With this in mind, we stress the importance of rigorously investigating the roles of and interactions between multiple genetic and non-genetic factors in the pathogenesis of PAD.

Functional genomics to prove variant pathogenicity

Due to the diversity of causal genes underlying PAD, an array of functional analyses is required for better and more rapid validation of variant pathogenicity, including VUS. Functional validation consists of determining a deleterious effect on gene function, e.g., cellular protein expression, complementation experiments, immune cell function and in vivo mouse models (Fig. 4). The amount of functional evidence required to prove causality depends on whether the variant is in a known (category 4) or newly identified (category 3) gene. Here, we will discuss strategies for functionally validating category 3 and 4 variants, using cases from the literature to illustrate these approaches.

Protein expression and complementation experiments

The minimum evidence required to prove variant pathogenicity is the demonstration that protein expression is altered, impaired, or absent (Fig. 4). To investigate this, an antibody to the protein of interest must be available, either for flow cytometry or western blot, and knowledge of the expression pattern of the encoded protein in different immune cells should be identified.²¹¹ Although proof of a deleterious effect on the encoded protein is strong evidence of pathogenicity, this is only sufficient where variants in the gene, resulting in absent protein expression, have been shown to be pathogenic. In patients with BTK variants, detection of the absence of circulating B cells, in combination with the absence of cytoplasmic BTK expression in monocytes,^212,213 would support a diagnosis of XLA. In contrast, in genes where multiple effects have been observed, despite normal protein expression, more in-depth analysis is required to prove causality. For example, CD70 is a transmembrane protein expressed on activation of B, T, and NK cells, and in patients with biallelic mutations in CD70, upregulation of protein in response to stimulation is impaired,^214–216 or CD70 binding to its receptor CD27 is abrogated despite normal CD70 expression on the cell surface.²¹⁴ Moreover, a genetic variant can have a negative effect on the expression of the encoded protein as well as a protein with which it interacts. For example, AR LOF variants in CD81 cause the absence of CD81 and CD19 on the B-cell surface as a result of intracellular retention of CD19 in the absence of CD81.⁸⁸ Furthermore, in LRBA deficiency, loss of LRBA expression causes increased degradation of CTLA-4 (cytotoxic T-lymphocyte-associated protein 4), resulting in impaired trafficking to the cell surface of Treg cells.²¹⁷ In these diseases, the functional defect is caused by absence of the interacting protein, and demonstration of altered expression of both proteins is required to validate pathogenicity.

For novel variants in known genes or newly identified genes, more intensive examination is required to prove a deleterious effect (Fig. 4). Generally, complementation assays should be utilized in functional validation, as protein expression alone may not wholly exclude abnormal protein function. In these studies, ideally a wild-type allele is transduced into primary cells or patient-derived cell lines such as B-lymphoblastoid cell lines (B-LCLs). Correction of the cellular phenotype upon complementation with the wild-type allele is considered crucial for confirmation of variant pathogenicity.^3,160 This is well demonstrated in CD81 deficiency, where the introduction of wild-type CD81 but not CD19 restored the expression of CD19 and CD81 on the B-cell surface, indicating that CD81 is required for CD19 expression.⁸⁸

These strategies are robust for proteins with known functions and where antibodies are available for detection. However, where antibodies are not available for the protein of interest different approaches need to be employed. One strategy is to monitor signaling pathway function, which requires knowledge of the role of the gene in signaling and the availability of a biochemical assay for testing (Fig. 4). The most commonly affected pathways in PAD are the PI3K and NFκB pathways (Fig. 3).^121,122 Therefore, the identification of novel variants in known causal genes or in newly identified genes within these pathways can be assessed. To accurately determine the pathogenicity of these variants, both resting state and post stimulation levels of phosphorylated common effector proteins, e.g., AKT and S6 for the PI3K pathway and p50 and IκBα for the NFκB pathway, should be assessed. NFκB pathway activity can be assessed in response to stimulation either by quantitation of NFκB effector proteins including p50 and IκBα in a western blot^134,142 or via measurement of their translocation to the nucleus by imaging flow cytometry in B or T cells.¹⁴²

Alternatively, levels of both phosphorylated AKT or S6 increase substantially upon stimulation in healthy controls providing baseline activity of the PI3K pathway, allowing for comparison with affected individuals. As demonstrated in PIK3CD GOF patients, static levels of phosphorylation of AKT and S6 upon stimulation confirm constitutive hyperactivity of this pathway.^145,148 In contrast, a decrease in total or phosphorylated protein, either of the encoded protein itself or other proteins in the signaling cascade, implies LOF of the encoded protein. Thus, these methods can distinguish between different mechanistic effects in a single causal gene. For example, biallelic LOF mutations in PIK3R1 and PIK3CD result in loss of p85α or p110δ protein expression, while AD LOF mutations in PIK3R1 or AD GOF in PIK3CD result in abrogated binding of p85α to p110δ, resulting in increased phosphorylation of AKT and S6.^{33,123–125,129,145,218} Therefore, the identification of reduced PI3K pathway activity as opposed to hyperactive PI3K pathway activity can help differentiate these diseases.

In conclusion, proof of a deleterious effect on protein expression and correction of the cellular phenotype by complementation studies is sufficient to prove pathogenicity for category 4 (likely pathogenic) variants (Fig. 4). However, for VUS genes (category 3), more extensive in vitro and in vivo analyses are required, which include studies to demonstrate that the variant has a deleterious effect on immune cell function (Fig. 4).

Impact of variant on immune cell function

Immunophenotypic analysis of the immune cell compartment

Quantification and characterization of B, T and innate cells can identify the impact of variants on the composition of the immune compartment and identify patterns that associate with different clinical phenotypes and genotypes, which may help direct genomic analysis in the future. This emphasizes that in-depth immunophenotyping should be undertaken in parallel with genomics. We recently showed that alterations in B, T and innate cells are associated with non-infectious complications in PAD.²¹⁹ Extended phenotyping past basic subsets can include expression of regulatory receptors, including inhibitory molecules PD-1 and CD57 demarking exhausted or senescent T cells, or the inclusion of tetramers to delineate antigen-specific memory cells. Exhaustion is defined by the upregulation of inhibitory molecules, including PD-1, whereby receptor ligation results in functional exhaustion, including reduced cytokine production.²²⁰ In contrast, cellular senescence characterized by upregulation of molecules including CD57 is known to correlate with reduced proliferative capacity and terminal differentiation.²²¹ Expansion of exhausted and/or senescent T cells is a hallmark of diseases in specific genetically defined PID. For example, expansions of senescent effector memory CD8⁺ T cells,^{126,148,151,219} in combination with transitional B cells,¹²⁵ central memory CD4⁺ T cells and T follicular cells,¹⁵⁰ are indicative of AD GOF PIK3CD variants.

MHC tetramers have traditionally been used to identify antigen-specific T cells.²²² In PID, these reagents have identified defects in the generation and immunophenotype of EBV-specific memory T cells, providing insight into the mechanisms underlying EBV susceptibility in a subset of patients.^214,219,223 Alternatively, if a protein has not been mapped for a particular pathogen or HLA allele, then immune cell stimulation with whole protein can be utilized to identify antigen-specific memory T cells.²²² More recently, fluorescently-labeled antigen tetramers have been utilized to assess the generation and immunophenotype of antigen-specific memory B cells in response to vaccination.^224,225 In patients receiving IgRT, it is impossible to accurately determine a vaccine response on the basis of serum IgG because antibodies generated by the patient are indistinguishable from donor antibodies in the circulation. In a proof-of-concept study, we recently used antigen tetramers to identify influenza-specific memory B cells pre- and post-vaccination. We showed that influenza-specific memory B cells were identifiable in all studied healthy individuals and PAD patients, albeit at lower numbers in patients both pre- and post-vaccination. Hence, the identification of influenza-specific memory B cells may represent an additional diagnostic test for PAD. With in-depth examination of the immunophenotype, this method will provide valuable information regarding the defects in antigen-specific memory B-cell generation in PAD patients.²²⁶

Immune cell function

In addition to quantitative changes, genetic variants may negatively impact immune cell activation, proliferation, and functionality.^88,95,125 Activation can be assessed either by assessing calcium influx or upregulation of activation markers in response to stimulation. Calcium influx after engagement of antigen receptors is required for immune cell activation. B-cell (anti-IgM) or T-cell (anti-CD3) specific stimulation triggers a sharp and rapid spike in intracellular calcium flux observed via flow cytometry. A blunted response to anti-IgM stimulation indicates defective BCR signaling and activation, as has been demonstrated in CD19 and CD81 deficiencies.^88,89,95 In contrast, impaired upregulation of activation markers, including CD86 on B cells and CD25 or CD69 on T cells, may be used as indirect markers of immune cell activation.^{148,214,227–229}

Cell proliferation occurs upon antigen recognition by immune cells. The capacity to proliferate in response to stimulation can be assessed by the incorporation of cell tracking dyes whereby the tracker dye signal is diluted with each subsequent cell division.²³⁰ Of relevance, diminished T-cell proliferation in response to recall antigens, but not mitogens including CD3, has been observed in CVID patients.²²⁷ This demonstrates that careful selection of stimuli is required to dissect the impact of each variant on immune cell functionality and hence its potential impact on disease pathogenesis.

Effector cell function differs substantially between cell types. B cells produce high-affinity antibodies, which are most easily quantified by Ig isotypes in serum.^231,232 In contrast, T-cell effector function comprises cytokine production or target cell lysis of infected or malignantly transformed cells.²²² The stimulation time length to determine T-cell effector function varies based on the cytokines assessed, the stimulus used and the readout. Intracellular cytokine production in response to non-specific stimuli, such as phorbol 12-myristate 13-acetate and ionomycin, or antigen-specific stimuli, including antigenic peptides, takes 6 h or up to 16 h if stimulated with whole protein antigens. The latter enables examination of the effector function of antigen-specific memory T cells. In contrast, quantification of secreted cytokines typically requires 48 h of stimulation.²²² CD8⁺ and CD4⁺ T-cell responses are distinct. IFN-γ, TNF-α, IL-2, granzyme B, perforin, and CD107a expression are routinely used to assess CD8⁺ T-cell cytokine secretion and cytotoxic capacity. Dysfunctional CD8⁺ T cells have been implicated in multiple PIDs, particularly in those predisposing patients to herpesvirus infection especially EBV.^{214,219,222,223,227,233} Alternatively, a panel of cytokines spanning those produced by different Th subsets (Th1, Th2, Th17, Treg) is used to determine whether a variant skews CD4⁺ T-cell differentiation towards a specific Th phenotype.²³⁴ Therefore, in-depth immunophenotyping and functional analyses can determine the role of a gene in immunobiology and the cellular mechanisms driving disease. This has the potential for the discovery of new cellular targets for therapy.

Collectively, proof of a deleterious effect on protein expression, immune cell function and correction of the cellular phenotype by complementation experiments is required to prove pathogenicity for category 3 (VUS) variants. Furthermore, if a variant impacts immune cell development rather than maturation or function, a murine model of disease showing recapitulation of the developmental defect is required to prove causality. Proof of causality can elevate the pathogenic categorization of a variant, thereby enabling a confirmatory diagnosis (Fig. 4), and expedite diagnosis on detection in future patients. Over time, this has the capacity to increase the diagnostic rate of PAD.

How genetic diagnosis can direct precision medicine

Molecular diagnosis of PID can dramatically impact patient management, shifting treatment from nonspecific therapies, including IgRT and prophylactic antibiotics, to targeted immunotherapeutics including monoclonal antibodies and biologics. The development of new biologics has been driven by the search for targetable molecules either to block immune regulatory pathways, such as PI3K, or to block T-cell responses dependent on the CTLA4 pathway as with abatacept. Therefore, identification of dysregulation in certain pathways in PID has seen repurposing of these previously approved therapeutics for use in the immunology clinic. Here, we describe examples of how molecular diagnosis has directed more targeted treatment in PIDs with the common feature of antibody deficiency.

Treatment of activated phosphoinositide 3-kinase δ (PI3Kδ) syndrome

Patients with LOF mutations in PIK3R1 and PTEN^122,129 or GOF mutations in PIK3CD^145,148 are susceptible to recurrent sinopulmonary infections, herpesvirus infections, autoimmunity, autoinflammation, lymphoproliferation, and lymphoma.^{9,122,145–147,219,235,236} Although both PIK3CD and PIK3R1 underlie activated phosphoinositide 3-kinase δ (PI3Kδ) syndrome (APDS), the similar clinical phenotype and shared constitutive hyperactivation of the PI3K–AKT–mTOR pathway in these diseases provides a rationale for targeted treatment with mTOR inhibitors (rapamycin) or specific p110δ inhibitors (leniolisib and nemiralisib). While rapamycin has been shown to limit lymphoproliferation of effector T cells,^148,236 treatment with Leniolisib decreases lymphoproliferation and corrects immunophenotype by increasing naive B-cell proportions, senescent T-cell proportions, and inflammatory cytokine levels, including IFN-γ and TNF, and reducing transitional B-cell frequencies.²³⁷ This promising study on six APDS patients will require follow-up in larger cohorts to confirm the findings. Treatment with Nemiralisib has shown clinical improvement in chronic obstructive pulmonary disorder²³⁸ and is currently being trialed for APDS patients with sinopulmonary infections and bronchiectasis (ClinicalTrials.gov Identifier: NCT02593539).

Treatment of CTLA-4 and LRBA deficiencies

CTLA-4 haploinsufficiency and LRBA deficiency both present with recurrent infections, severe autoimmunity and decreased B cells, although differences in clinical presentation exist, with variable penetrance seen within pedigrees.^{198,217,239–244} CTLA-4 expression is dysregulated in both diseases, with decreased expression of CTLA-4 identified in CTLA-4 haploinsufficiency²⁴¹ and in LRBA deficiency, where the absence of LRBA protein, an intracellular chaperone of CTLA-4 to endosomes for recycling, results in increased CTLA-4 degradation.²¹⁷ Reduced expression of CTLA-4, an immune checkpoint, provided rationale for the use of CTLA4-IgG fusion proteins Abatacept and Belatacept. These therapeutics have been shown to restore the checkpoint imbalance, thus reducing clinical symptoms under both conditions.^{198,240,245,246}

Treatment of STAT1 GOF

AD variants in signal transducer and activator of transcription (STAT)1 GOF result in chronic mucocutaneous candidiasis (CMC), bacterial and viral infections, autoimmunity and immune dysregulation. This clinical phenotype is driven by reductions in T helper 17 (Th17) and in some patients, memory B-cell proportions.^247–252 STAT1 functions downstream of interferon receptors to drive cell proliferation, differentiation and survival and to inhibit STAT3 function.²⁵³ STAT1 GOF mutations lead to exaggeration of STAT3 inhibition and subsequent impairment of Th17 immunity.^251,253 Jakinibs are small molecule biologics that inhibit JAK protein signaling and inhibit STAT-mediated intracellular signals. Ruxolitinib and Baracitinib, which target JAK1 and JAK2, have been successfully used in some cases of STAT1 GOF to treat CMC and autoimmune manifestations.^254–256

Hematopoietic stem cell transplantation as definitive treatment of PAD

Although small molecule biologics and monoclonal antibodies have improved the prognosis of some PID patient groups, hematopoietic stem cell transplantation (HSCT) is still the only curative treatment for PID.

In a subgroup of PAD patients, IgRT and targeted immunotherapy are insufficient to mitigate the early mortality and high mortality rates associated with disease. As such, immunomodulatory drugs may provide a bridge to more high-risk definitive treatments, including HSCT and gene therapy. Although HSCT is the standard of care for patients with defined genetic defects, including severe combined immunodeficiency (SCID), a low number of CVID patients (<40) have undergone HSCT to date,^257–259 likely due to the heterogeneity of the clinical, immunological and genetic phenotypes. In these studies, transplantation was indicated in all patients on the basis of lymphoma, treatment-refractory severe infections, complex immune dysregulation, or impending organ failure.^257–259

Significant advances in transplantation medicine include the development of low intensity conditioning regimens, alternative donor stem cell sources including TCR αβ-CD19-depleted cells, and improved prophylaxis for graft versus host disease (GVHD).^260,261 Together with better antiviral therapeutics, including therapeutic administration of virus-specific T cells,^259,262 these improvements have dramatically reduced engraftment and GVHD rates and treatment-related mortality rates.²⁶³ However, despite these advances, post-transplant mortality rates in CVID are high (up to 52% vs 20% for SCID).^258,264,265 Poorer outcomes are associated with a positive family history of disease, pre-existing infections, and organ involvement (liver, renal, lung, or granulomatous).^258,259 In addition, as PAD patients typically present later in life, the lower engraftment rate is partially due to age-related thymic damage and involution, which delay thymic-dependent immune reconstitution, and alongside increased infection-induced organ damage, contribute to higher mortality in older recipients.^260,261 Thus, it has been suggested that transplantation should be restricted to PAD patients where the specific defect underlying disease is known and where clear evidence exists that B and/or T-cell engraftment will be corrective.²⁵⁷ Timing will ultimately be crucial in treating patients before the onset of end-stage organ damage to improve treatment outcomes.^257,258 Furthermore, larger multicenter studies are required to determine the optimal timing for HSCT, optimal patient selection for transplantation, and the level of chimerism required for a successful outcome.

In recent decades, genetics and genomics have significantly impacted the care of PAD patients. Importantly, those patients initially presumed to suffer from PAD, but after subsequent reclassification following genetic diagnosis, have received more relevant treatment including HSCT for CID. In multiple cases, molecular diagnosis has provided a rationale for targeted treatment of the affected pathways with new biologicals or small molecule inhibitors.^{198,237,240,245,246,254–256} A significant increase in the diagnostic rate, particularly of PAD, is required for personalized medicine to be applicable for the majority rather than the minority of patients.

Future directions

Over the next five to ten years, improved strategies are required to increase the diagnostic rate of PAD, with the potential to dramatically reduce the early mortality and high morbidity rates associated with this disease.

To advance diagnostic strategies in PAD, research should focus on the identification and functional validation of new causal genes as well as novel variants in existing disease-causing genes. This will require a multidisciplinary team, improved variant identification pipelines, robust testing of functional assays and design of a pipeline to enable a quick turnaround time for functional validation of variants. For these efforts to be successful, close collaboration between geneticists, clinical immunologists and scientists with expertise in functional genomics, transcriptomics, epigenetics, and bioinformatics will be required. Interactions between these core personnel are required to clinically validate multi-omics assays, ascertain how to integrate these analyses for better understanding of the genetic etiology of PAD and how or if it is possible to embed these analyses into routine diagnostics. Furthermore, resolution of the conundrum as to whether common variants represent biomarkers for NIC development will determine whether these variants should be clinically reported.

The role of epigenetic modifiers within the immune system remains poorly understood for primary immunodeficiencies, although some single gene variants, such as INO80, may drive pathogenicity due to alterations in epigenetics.^266,267 In addition, impaired methylation in genes driving the B-cell phenotype and function has been identified in CVID.²⁶⁸ Further understanding of these epigenetic changes may also potentially contribute to understanding the heterogeneity in certain PADs. These issues may be poorly understood in part due to the complexities and tissue-specific methylation characteristics. Future developments that may enable cell-free deconvolution of methylation states,²⁶⁹ if performed in tandem with in-depth immunophenotype studies, may provide further insights into the penetrance, heterogeneity, and severity of PAD. However, these investigations will require collaboration with highly skilled experts to enable design of experiments to gain understanding of the role of epigenetics in PID and whether these modifications represent therapeutic targets of disease.

Advances in other omics technologies, including unbiased cluster analysis of plasma proteomics and whole blood, have enabled the identification of proteomic and signaling pathway profiles associated with PAD. These methodologies have the potential to identify critical signaling and regulatory pathways impacted in PAD patients as a group. However, their utility on a patient-to-patient basis is unclear and will require further investigation in a research setting.^71,270

In conclusion, we believe that a broader view of the current genetics and immunologic diagnostic approach is required to expedite genetic diagnosis. It is becoming clearer that broader identification of rare (small and gross lesions), and to a lesser extent, common variants are required to better diagnose and treat patients. However, the reliance on WES alone is unlikely to be sufficient to meet this requirement, and as such a switch to WGS will be needed. Irrespective of the genomics methodology, improved functional validation is required to prove or disprove the relevance of VUS, as well as to better understand the contributions of common variants and non-genetic variants to disease pathogenesis. These insights have the potential to dramatically increase the diagnostic rate of PAD, identify biomarkers for NIC development and actualize personalized medicine for more PAD patients. Together, these strategies will hopefully improve health and curtail the early mortality and high morbidity currently associated with disease-related complications.

Author contributions

E.S.J.E. and M.C.v.Z. conceptualized and wrote the manuscript. J.J.B., S.O., and R.E.O.H. provided critical feedback and commented on manuscript drafts. All authors approved the final version.

Competing interests

The authors declare no competing interests.